3 replies to this topic

[Nicholas Christopher]

Anyone have any personal insight on the efficiency or not of
Orthogonal Illumination Maps for lighting the models in a scene
containing more than one model.

This technique requires 6 render passes per model , by using 6 pre-calculated textures
essentially a cube map, just for lighting.

If the texture is no larger that lets say 64x64 or 32x32 .

[Nicholas Christopher]

Just reflecting on this OIM method.

I figure that in most cases when rendering tri models in a game or otherwise 
where speed is a factor, models are rendered nicely but not correctly 
no matter what light model you use. You can correct me if I'm wrong.

And the OIM method is no different in this respect.
I just thought of one scenario where you could have a pinhole where
light can pass through. This does not mean the light will hit any part of
the model (in reality) , but if this pinhole appears on the maps it will appear
to be brighter on the model.

Just trying to measure up this technique, beforehand.

[Goz]

If im looking at the right thing. Its just per-pixel DP3 lighting performed on cards that can't support pixel shaders.

I'd stick with pixel shader based per-pixel if i was you ...

[Nicholas Christopher]

Hi Goz.

I understand your DOT3 per-pix rendered for every light source whose attenuation reaches the subject(model).

But what if you have more than 6 lights, + GI factors of light eminating through
doorways , windows or any others. In my case this is the norm, with easily 100
or more light sources.

Would it not be more efficient to make a form of static lit cubemap (OIM's) especially
in the case of static models (which are always more numerous than dynamic models)?